8

Yes, F0 (the fundamental frequency) is the acoustic correlate of pitch (which is a perceptual concept). The fundamental frequency F0 is also the first harmonic H1 of the sound. If F0 is 100 Hz, the second harmonic H2 would be at 200 Hz, the third H3 at 300 Hz, the fourth H4 at 400 Hz, and so on. Vowel formants are located at different harmonics depending on ...


5

While there are some context models that are unordered (Latent Semantic Analysis, some semantic word-vector approaches), that's not what you seem to be talking about. Instead, I think the critical point is that you're describing the n-grams incorrectly. An n-gram model is usually ordered, so that the whole sequence "three blind mice" is a trigram, and "blind ...


3

There is a limited sense in which F0 (which is the acoustic property perceived as pitch) and F1-F5 are not independent, which is that if you have a tiny larynx (high F0), given the nature of human anatomy you will not have a really long tube (which determines formant frequencies). Another dependency is that resonance frequency can always be meaningfully ...


3

Hope it is not too late. Both ppl and ppl1 are normalized according http://www.speech.sri.com/projects/srilm/manpages/srilm-faq.7.html , which is called entropy rate.


2

For parsing: if the word is already in your 'dictionary' then treat it as a 'unigram'; but if it is new to you then treat it as an 'ngram'. If you are doing semantic statistics on a corpus then you may need to do both.


2

(Disclaimer: I don't have much background in linguistics) I think - if you are relating to agglutinantive and not polysynthetic languges - it would depend on what you are trying to build and the language too. But, Staatspolizei (Secret State Police) may occur around words like- The (whereas polizei would require and adjective before e.g. The Secret-...


2

There are language resources called wordnets that represent graphs of hypernymy and hyponymy and synsets. The prototypical example of such a resource is the Princeton WordNet for the English language that is available under a free licence. There is another type of language resource called ontology. Typically, ontologies are constructed for particular ...


2

For formal English anyway you can parse with spaCy and then iterate through the tagged tokens looking for a finite verb (VerbForm=Fin, as opposed to Ger or Inf). See https://spacy.io/usage/linguistic-features#rule-based-morphology


2

I've seen one agent-based model recently: van Trijp, R. (2013), Linguistic Assessment Criteria for Explaining Language Change: A Case Study on Syncretism in German Definite Articles. This paper tries to explain the diachronic change of a small part of German grammar by having agents play "language games" (using Fluid Construction Grammar).


2

In my experience, speech-recognition systems do quite a bit of preprocessing on the signal before trying to interpret it; part of that is getting rid of the noise. The key is, most speech signals are periodic: the vocal folds generate a sawtooth-esque wave, and the vocal tract then applies a filter to it. So all voiced sounds(*) have a very predictable ...


1

Robust speech recognition is a huge area with lots of research papers. You can start from a textbook by Microsoft: Robust Automatic Speech Recognition


1

As I understand the terms, the distribution of a VP would be the X and Y in a constituent [ X VP Y ]. However, affixes associated with auxiliary verbs might count as part of the X rather than part of the VP (looking before the affix-hopping transformation, that is). The list of things you call environments seem to me rather to be descriptions of verbs or ...


1

Try Open source Julius Decoder: https://github.com/julius-speech/julius


1

Here is one from the LINDAT/CLARIN repository (for a specialised domain, but Open Data): ATCC: Pronunciation lexicon and n-gram counts for ASR module


1

I think http://www.collocates.info/ will have the data you need. Note, the full processed list is not free but is not outrageously expensive.


1

GT is SRILM's default. In fact, I think using -addsmooth 0 just gives you default GT smoothing (unfortunately?). To directly use GT discounting, simply include no discounting argument. The number at the end of the discount argument (-kndiscount1, etc.) is the n-gram order to apply that discounting method to. Using -kndiscount alone tells ngram-count to use ...


Only top voted, non community-wiki answers of a minimum length are eligible